ORIGINAL PAPER

O. R. Crasta á W. W. Xu á D. T. RosenowJ. Mullet á H. T. Nguyen

Mapping of post-¯owering drought resistance traits in grain sorghum:association between QTLs in¯uencing premature senescenceand maturity

Received: 10 October 1998 /Accepted: 12 July 1999

Abstract The identi®cation of genetic factors underlyingthe complex responses of plants to drought stress providesa solid basis for improving drought resistance. The stay-green character in sorghum (Sorghum bicolor L.Moench)is a post-¯owering drought resistance trait, which makesplants resistant to premature senescence under droughtstress during the grain®lling stage. The objective of thisstudy was to identify quantitative trait loci (QTLs) thatcontrol premature senescence and maturity traits, and toinvestigate their association under post-¯owering droughtstress in grain sorghum. A genetic linkage map was de-veloped using a set of recombinant inbred lines (RILs)obtained from the cross B35 ´ Tx430, which were scoredfor 142 restriction fragment length polymorphism(RFLP) markers. The RILs and their parental lines wereevaluated for post-¯owering drought resistance and ma-turity in four environments. Simple interval mappingidenti®ed seven stay-greenQTLs and twomaturity QTLs.Three major stay-green QTLs (SGA, SGD and SGG)contributed to 42% of the phenotypic variability (LOD9.0) and four minor QTLs (SGB, SGI.1, SGI.2, and SGJ)signi®cantly contributed to an additional 25% of thephenotypic variability in stay-green ratings. OnematurityQTL (DFB) alone contributed to 40% of the phenotypic

variability (LOD 10.0), while the second QTL (DFG)signi®cantly contributed to an additional 17% of thephenotypic variability (LOD 4.9). Composite intervalmapping con®rmed the above results with an additionalanalysis of the QTL ´ Environment interaction. Withheritability estimates of 0.72 for stay-green and 0.90 formaturity, the identi®ed QTLs explained about 90% and63% of genetic variability for stay-green and maturitytraits, respectively. Although stay-green ratings weresigni®cantly correlated (r � 0.22, P £ 0.05) with matu-rity, six of the seven stay-greenQTLswere independent ofthe QTLs in¯uencing maturity. Similarly, one maturityQTL (DFB) was independent of the stay-green QTLs.One stay-green QTL (SGG), however, mapped in the vi-cinity of a maturity QTL (DFG), and all markers in thevicinity of the independent maturity QTL (DFB) weresigni®cantly (P £ 0.1) correlated with stay-green ratings,confounding the phenotyping of stay-green. The molec-ular genetic analysis of the QTLs in¯uencing stay-greenandmaturity, together with the association between thesetwo inversely related traits, provides a basis for furtherstudy of the underlying physiological mechanisms anddemonstrates the possibility of improving drought resis-tance in plants by pyramiding the favorable QTLs.

Key words Sorghum bicolor (L) á Drought resistance áQuantitative trait loci (QTLs) á Trait-based QTLpyramiding

Introduction

Abiotic stress factors, of which drought and high tem-perature are the major ones, are considered to be themajor cause (71%) of yield reductions in crop plants(Boyer 1982). More than 80% of the sorghum in the USis grown under non-irrigated conditions, where water isthe major limiting factor for yield. Despite the majorresearch emphasis during the last two decades on im-proving drought resistance in sorghum (Rosenow et al.1983), progress in this regard has been slow.

Mol Gen Genet (1999) 262: 579±588 Ó Springer-Verlag 1999

Communicated by R. Hagemann

O. R. Crasta á W. W. Xu á H. T. Nguyen (&)Plant Molecular Genetics Laboratory,Department of Plant and Soil Sciences,Texas Tech University, Lubbock, TX 79409, USAE-mail: bwhtn@ttacs.ttu.eduTel.:+1-806-742-1622; Fax: +1-806-7420775

O. R. CrastaCuraGen Corporation, New Haven, CT 06511, USA

W. W. Xu á D. T. RosenowTexas A and M University Agricultural Researchand Extension Center, Box 219, Lubbock, TX 79401, USA

J. MulletDepartment of Biochemistry and Biophysics,Texas A and M University, College Station, TX 77843, USA

Strategies for crop improvement with respect todrought resistance include the identi®cation and selec-tion of traits that, at least partly, contribute to improvedperformance of the crop under drought conditions. Thistrait-based crop improvement strategy allows selectiveaccumulation of the traits that contribute to droughtresistance for a speci®c target environment (Blum 1983;Rosenow et al. 1983; Ludlow and Muchow 1990).However, the success of this approach is limited by thedi�culty experienced in identi®cation of genotypes forseveral traits, due to lack of proper control of the in-tensity and timing of stress. Success is further reduced bythe high cost and the large amount of labor involved inconducting such multi-location experiments.

In grain sorghum, the ability to resist premature se-nescence due to post-¯owering drought stress is termedthe ``stay-green'' trait (Rosenow et al. 1983). Plants withthe stay-green trait resist premature plant and leaf death,develop grain normally, and resist charcoal rot andlodging when exposed to moisture stress during the latestages of grain development (Rosenow and Clark 1981;Rosenow et al. 1983; Rosenow 1984; Tenkouano et al.1993; Walulu et al. 1994). The stay-green phenomenonhas been extensively studied in plants, motivated byseveral economic incentives (Noode n 1988a; Thomasand Smart 1993; Bleecker and Patterson 1997). Recentstudies have demonstrated that leaf senescence is a ge-netically programmed phenomenon (Oh et al. 1997) andthere is growing interest in studying the molecularmechanisms that underlie this process (Buchanan-Wollaston and Ainsworth 1997; Gri�ths et al. 1997;John et al. 1997; Kleber-Janke and Krupinska 1997;Lers et al. 1998).

Crop plants have been selected for early maturityunder terminal drought stress conditions, which in-creases the probability of encountering favorable mois-ture conditions during the more critical reproductivephase (Ludlow and Muchow 1990). While this pro-grammed completion of the life cycle under conditionsof severe terminal drought stress ensures e�cienttranslocation of nutrients to the sink, premature senes-cence a�ects the assimilatory capacity and the durationof the assimilatory phase, resulting in drastic reductionin grain ®lling. In this study we have identi®ed theMendelian factors in¯uencing stay-green and early ma-turity, and investigated the phenotypic and genetic as-sociation between these two seemingly inversely relatedtraits. Understanding of the genetic association betweenthese two traits facilitates pyramiding of QTLs for im-provement of drought resistance in crop plants.

Materials and methods

Plant material

Two genotypes, B35 and Tx430 were selected because they showdistinct di�erences in drought response and yield potential. B35 hasoutstanding post-¯owering (stay-green) drought resistance. How-ever, it has a relatively low yield potential. Tx430 is a high-yielding

line with exceptionally wide adaptation and is used worldwide inbreeding programs. Tx430, however, is susceptible to post-¯ower-ing drought stress. The F1 lines obtained from the crossB35 ´ Tx430 were selfed in all successive generations to produceone F6 line from each of the 96 F2 plants. The seeds from each ofthe 96 F6 lines were bulked and used for phenotyping and geno-typing as 96 F6:7 recombinant inbred lines (RILs).

B35, Tx430, and the 96 F6:7 RILs were grown in ®eld experi-ments under post-¯owering drought stress conditions (stress) infour environments: Lubbock, Texas during 1993 (ENV1) and 1994(ENV2) and Halfway, Texas during 1993 (ENV3) and 1994(ENV4). Three irrigations were applied during the pre-¯oweringgrowth period to minimize the pre-¯owering water de®cit. Theexperiments were carried out in a randomized complete block de-sign with three replications. Each plot consisted of one row ofplants, 4.9 m long, with a 1.0 m row-spacing. Each replicationcontained random repetitions of the parents, B35 and Tx430, oncefor every 10 rows of RILs.

Phenotyping

Plots were evaluated for the stay-green trait near the end of thelinear grain-®ll period. At each environment, visual ratings wererecorded for expression of the stay-green trait (scores ranged from1 to 5 based on the degree of leaf and plant death; score 1 repre-sents no senescenced leaves and score 5 representing all senescedleaves). Chlorophyll index readings were taken with SPAD-502chlorophyll meter (Spectrum Technologies Inc), and were recordedat the same time only during the 1994 growing season to back-upthe visual ratings of stay-green. Maturity ratings were recorded asthe number of days from planting to 50% anthesis. Statisticalanalysis was performed using the Proc GLM (SAS Institute 1989)procedure to evaluate the parents and RILs for genetic variation instay-green and chlorophyll index assuming the ®xed-e�ects model(Model I). Maturity ratings followed the normal distribution, whilelog transformation was done on the stay-green ratings to ®t thenormality assumption. Broad-sense heritability (H) estimates werecalculated on a family-mean basis from pooled analysis as the ratioof genetic variance (r2g) to phenotypic variance (r2ph) (Fehr 1987).However, the random-e�ects model (Model II) was assumed forcalculation of heritability estimates.

Genotyping

Genomic DNA was isolated from the parental lines (B35 andTx430) and the 96 RILs (based on Saghai-Maroof et al. 1984).Southern blots were prepared by digesting 10 lg of DNA using fourrestriction endonucleases (EcoRI, EcoRV, BamHI, and HindIII;Promega, Madison, Wis.) and, following electrophoresis, transfer-ring the DNA to Hybond N+ membrane (Amersham Life Sci-ences, Arlington Heights, Ill.) as recommended by the membranemanufacturer.

The parental lines were screened for restriction fragment lengthpolymorphism (RFLP) using sorghum genomic clones (obtainedfrom Drs. G. Hart and A. Paterson; Texas A and M University),maize cDNA and genomic clones (Univ. of Missouri, Columbia),clones mapped on rice and wheat (Cornell University, N.Y.), andother cDNA clones available in our collection. Labeling of theprobes was done either by PCR-random primer labeling (BIOSTag-it-kit) or by oligolabeling (Feinberg and Vogelstein 1984). Theprobes that were polymorphic between the parental lines were usedfor hybridization with DNA from the 96 RILs and the parentallines. The RILs were scored for markers as `A' and `B' for presenceof the parental band of the female parent (B35) and male parent(Tx430), respectively, `H' for heterozygotes, `C' for non-femaleparent, `D' for non-male parent, and `-' for missing and non-pa-rental markers.

The RFLP linkage map was developed using MAPMAKER(Lander et al. 1987). A LOD score of 3.0 was used to establishlinkage between markers. The level of heterozygosity within each

580

line ranged from 0 to 24%, with an average of 11%, among co-dominant markers. Therefore the MAPMAKER option ``F2 in-tercross'' instead of ``RIL'' was used for developing the geneticlinkage map. However, the ``RIL'' option, which requires completehomozygosity, was tried and the resultant map order did notchange, although the map distances were a�ected because the he-terozygotes were treated as missing values. Genome compositionwas estimated based on marker genotypes and the map distancesbetween markers as described by Paterson et al. (1991).

QTL analyses

Simple interval mapping (Paterson et al. 1988) was applied to de-termine the chromosomal locations of putative QTLs in¯uencingstay-green and maturity traits, while the composite interval map-ping (Jansen 1993) method was applied to evaluate theQTL ´ Environment interactions. Simple interval mapping wasdone using the MAPMAKER/QTL program (Lander and Botstein1989). The pooled means of the traits across four environmentswere used for mapping QTLs. A LOD threshold of 1.3 was initiallyused to identify ``putative'' QTLs (the LOD thresholds are oftenlowered to identify ``putative'' QTLs in¯uencing complex traitswith large G ´ E interactions; Stuber et al. 1992). The putativeQTLs were further evaluated in a step-wise manner using LODscore as selection criterion. All putative QTLs with a minimumLOD increment of 2.0 were considered as candidate QTLs.

Composite interval mapping was done using PLABQTL soft-ware (Utz and Melchinger 1996). The pooled means of traits wereused for mapping QTLs and the means from individual environ-ments were used to estimate the QTL ´ Environment interaction.Initially, markers were selected as cofactors by stepwise regressionanalysis with a stringent F-enter and F-drop value of 5.0 each.These selected markers ± except those on the current chromosome ±were used as co-factors (``cov /-'' command). A LOD threshold of2.0 was used to determine the presence of the candidate QTL. FinalQTL analysis was done by using the markers adjacent to the can-didate QTLs as cofactors with a LOD threshold of 2.0.

Results

Genotyping

Out of the 142 RFLP markers used for mapping, 128markers were mapped to 14 linkage groups (LGs)spanning a total of 1602 centiMorgans (cM in the Ko-sambi function) (Fig. 1). The average distance betweenadjacent markers was 14.1 cM. Approximately 51% ofthe intervals between adjacent markers were smallerthan 20 cM and 26% were in the range from 20 to30 cM. Sixteen markers were scored as dominantmarkers, while 112 were scored as codominant markers.Average heterozygosity among markers within an indi-vidual ranged from 0 to 24%, with an average of 11%.Distorted segregation of homozygous alleles (expectedratio 1:1, excluding heterozygotes) occurred (P £ 0.05)at 17 RFLP loci.

On average, 43.5% of the genome was homozygousfor B35 alleles, 46.7% of the genome was homozygousfor Tx430 alleles, while 9.8% of the genome was het-erozygous (Fig. 2). The v2 test against the expected ratioof 48.4:3.1:48.4 for aa:ab:bb alleles for generation F6(since the RILs were grown from bulked seeds from F6lines) was signi®cant (v2 � 14.8, k � 2, P £ 0.0001).The distortion of the segregation ratio was solely due to

the increased proportion of heterozygotes compared tohomozygotes, as the proportion of homozygotes(43.5:46.7 for aa:bb) was within the expected limits for1:1 (v2 � 0.79, k � 1, P � 0.37). The segregation ratiowas equivalent to the expected ratio of 46.9:6.2:46.9 foraa:ab:bb alleles in the F5 generation (v2 � 2.26, k � 2,P � 0.32), suggesting that the development of RILsmay have been slowed down due to bulking of the seedsin the F3 generation and may not re¯ect strict single-seed descent breeding, which favors the taller and latermaturing heterozygous plants. This notion is strength-ened by the Poisson-type distribution of heterozygousgenome, where median and mode are more representa-tive measures of central tendency than mean (Fig. 2).

Phenotyping

The stay-green ratings were signi®cantly lower (greener)in B35 (pooled average of 1.8) than in Tx430 (pooledaverage of 3.4) at all four environments (Fig. 3). Thechlorophyll index values for B35 (48.5) were also sig-ni®cantly higher compared to Tx430 (28.6). However, infull-irrigation environments, the chlorophyll index val-ues for the two lines were comparable (data not shown).Stay-green ratings among the RILs ranged from 1.5 to4.3, with an average of 2.6. The heritability estimate ofthe stay-green trait was 0.72. The stay-green trait washighly correlated with chlorophyll index (r � )0.83)(Table 1), indicating that visual ratings were reliable inestimating the degree of premature senescence.

The maturity ratings among the RILs ranged from 58to 76 with an average of 64 (Fig. 3). The heritabilityestimate for maturity was 0.90. Tx430 ¯owered about 6days earlier than B35 in all four environments. Whenpooled across the environments, maturity ratings werenegatively correlated with stay-green ratings among theRILs (r � )0.22, P £ 0.05), but were not correlatedwith the chlorophyll index values (r � 0.15) (Table 1).

Mapping of the stay-green trait. Simple intervalmapping identi®ed a total of seven QTLs that in¯uencethe stay-green trait (Table 2). Together these sevenQTLs contributed to 66.5% of the phenotypic variabil-ity, with a cumulative LOD score of 20.3. Initially, threemajor QTLs (SGA, SGD and SGG) in¯uencing the stay-green trait were detected, each contributing to more than10% of the phenotypic variability with LOD scores ofmore than 2.0 (Table 3). Together these three majorQTLs contributed to 41.2% of the phenotypic variabil-ity with a cumulative LOD score of 9.0. The residualstay-green phenotypic variability, after ®xing these threemajor QTLs, was used to identify four minor QTLs,each signi®cantly (LOD score of ³2) contributing to theresidual phenotypic variability (2% or more). Althoughone more QTL was initially observed on LG A betweenthe markers TXS422 and TXS1927, with a LOD score of2.3, it did not contribute signi®cantly (LOD score of 2 ormore) to the residual phenotypic variability after ®xingthe three major QTLs (data not shown).

581

582

Composite interval mapping identi®ed the same sevenQTLs discovered by simple interval mapping (Table 3).In all three major QTLs (SGA, SGD and SGG) and twominor QTLs (SGI.2 and SGJ), alleles from B35 con-tributed to an improvement (lower score) in the stay-green rating, while in two of the minor QTLs (SGB andSGI.1) alleles from Tx430 improved the stay-green rat-ing. The estimate of the contribution to phenotypicvariability by the seven QTLs (63.9% � 6.1) in com-posite interval mapping was comparable to that esti-mated by simple interval mapping (66.5%). Each of themajor QTLs contributed more than 20% of the pheno-typic variability. The regression coe�cients of all threemajor QTLs for the pooled values of the stay-green traitwere highly signi®cant (LOD score greater than 5). Theregression coe�cients of the two major QTLs (SGA andSGD) were signi®cant in all four environments, whilethat of the third major QTL (SGG) and two minorQTLs (SGB and SGI.2) were signi®cant in at least threeenvironments. The regression coe�cients of the QTLsSGI.1 and SGJ were signi®cant in two and one envi-ronments, respectively (Table 4).

Mapping of maturity trait

Two QTLs (DFB and DFG) in¯uencing maturity wereidenti®ed by both simple and composite interval map-ping methods (Tables 2 and 3). In both the QTLs, allelesfrom Tx430 contributed to early maturity. One QTL(DFB) alone contributed more than 40% of the phe-notypic variability. The regression coe�cients for boththe QTLs were signi®cant in all four environments(Table 4). Initially, the simple interval mapping methodidenti®ed one more putative QTL in¯uencing maturityon LG B (interval TXS1684-TXS1694) with a LOD

score of 2.5. Similarly, composite interval mappingidenti®ed the same QTL (interval pSB355-TXS1684 witha LOD of 3.4) and two QTLs on LG G (intervalsTXS1129-CDO20 with a LOD of 4.1 and CDO20-UMC27 with a LOD of 6.0). However, none of theseQTLs signi®cantly contributed to the residual pheno-

Fig. 1 RFLP linkage map of sorghum showing the QTLs in¯uencingstay-green and maturity traits under post-¯owering drought stressenvironment. The map was developed using the F6:7 RIL populationof B35 ´ Tx430. The map distances are given in centiMorgans(Kosambi function)

b

Fig. 2 Frequency distribution of the genome composition of RILsestimated based on marker genotypes (as described by Paterson et al.1991)

Fig. 3 Frequency distribution of the stay-green and maturity ratings(pooled averages over four environments) in the RIL population ofB35 ´ Tx430. The average values of the parental lines and the RILsare indicated

583

Table

1Sim

ple

correlationsbetweenvariousphenotypic

meansofRIL

s

Trait

EnvironmentStay-green

aMaturity

aChlorophyllindex

a

ENV1

ENV2

ENV3

ENV4

Pooled

ENV1

ENV2

ENV3

ENV4

Pooled

ENV2

ENV4

Stay-green

ENV2

0.56

(***,94)

ENV3

0.62

(***,95)

0.38

(***,94)

ENV4

0.40

(***,90)

0.41

(***,90)

0.46

(***,90)

Pooled

0.78

(***,90)

0.84

(***,90)

0.71

(***,90)

0.75

(***,90)

Maturity

ENV1

)0.32

(**,95)

)0.08

(ns,94)

)0.37

(***,95)

)0.09

(ns,90)

)0.21

(ns,90)

ENV2

)0.25

(*,94)

)0.04

(ns,94)

)0.35

(***,94

)0.02

(ns,90)

)0.13

(ns,90)

0.70

(***,94)

ENV3

)0.33

(**,95)

)0.13

(ns,94)

)0.39

(***,96)

)0.05

(ns,90)

)0.23

(*,90)

0.84

(***,95)

0.79

(***,94)

ENV4

)0.27

(*,90)

)0.11

(ns,90)

)0.40

(***,90)

)0.07

(ns,90)

)0.22

(*,90)

0.76

(***,90)

0.77

(***,90)

0.88

(***,90)

Pooled

)0.31

(**,90)

)0.09

(ns,90)

)0.42

(***,90)

)0.06

(ns,90)

)0.22

(*,90)

0.87

(***,90)

0.87

(***,90)

0.96

(***,90)

0.96

(***,90)

Chlorophyll

Index

ENV2

)0.47

(***,94)

)0.83

(***,94)

)0.29

(**,94)

)0.40

(***,90)

)0.71

(***,90)

0.07

(ns,94)

0.05

(ns,94)

0.08

(ns,94)

0.06

(ns,90)

0.04

(ns,90)

ENV4

)0.41

(***,90)

)0.34

(**,90)

)0.52

(***,90)

)0.68

(***,90)

)0.61

(***,90)

0.19

(ns,90)

0.19

(ns,90)

0.19

(ns,90)

0.20

(ns,90)

0.21

(*,90)

0.33

Pooled

)0.57

(***,92)

)0.78

(***,92)

)0.53

(***,92)

)0.60

(***,90)

)0.83

(***,90)

0.12

(ns,92)

0.11

(ns,92)

0.16

(ns,92)

0.15

(ns,90)

0.15

(ns,90)

0.84

(***,92)

0.66

(***,90)

aThevalues

given

are

Pearsoncorrelationcoe�

cients.Thelevelofsigni®cance

isgiven

inparentheses

(***,P

£0.001;**,P

£0.0;*,P

£0.0;ns,P

³0.05),followed

bythesamplesize

584

typic variability once the major QTLs were ®xed (insimple interval mapping) or used as cofactors (in com-posite interval mapping).

Discussion

Genotyping

Markers common to our linkage map (Fig. 1) andother published maps (Chittenden et al. 1994; Xu et al.1994) were used to compare the LGs identi®ed. The®rst 12 LGs (A to J) corresponded to the ten largestLGs reported by Xu et al. (1994). Eleven linkage as-sociations (A, B, C, D2, E, F, G, H, I1, I2 and J)corresponded to the ten complete linkage associations(A, I, F, B, J, D, C, E, G, G, and H, respectively)reported by Chittenden et al. (1994). The linkage order

of the common single-copy RFLP markers agreed withthe maps reported by Xu et al. (1994) and Chittendenet al. (1994). However, some of the RFLP loci detectedwith multicopy markers mapped to a di�erent location,suggesting polymorphism of the clone from a di�erentlocation. This map also serves to compare the twoabove-mentioned sorghum RFLP maps published ear-lier (Chittenden et al. 1994; Xu et al. 1994). The lowernumbers of markers on chromosomes C, E and Icompared to the above-mentioned maps are probablydue to low levels of polymorphism between the parentsin these regions.

Phenotyping

The phenotypic values for stay-green rating (Fig. 3)con®rmed the capacity of B35 to resist premature se-

Table 2 Simple interval map-ping of QTLs in¯uencing stay-green and maturity traits ingrain sorghum

Trait/stepa QTL LG Marker Position CumulativeLOD

CumulativeR2 (%)

Stay greenStep1 SGA A TXS307 23.3

SGD D2 TXS1537 3.1SGG G UMC27 4.8 9.0 41.2

Step2 SGI.2 I2 TXS1541 0 12.8 54.1Step3 SGI.1 I1 pSB134.2 0.4 15.4 59Step4 SGB B pSB115 4.2 18.2 64.5Step5 SGJ J TXS713 27.8 20.3 66.5

MaturityStep1 DFB B TXS1299 0.1

DFG G UMC27 15.8 14.9 57.4

a The predicted mean values for the traits in the parental populations are: for stay-green, 1.95 and 3.07in B35 and Tx430, respectively, and for maturity 69.2 (B35) and 60.0 (Tx430). For details of ratingprocedures, see Materials and methods

Table 3 Comparison of simple and composite interval mapping of QTLs in¯uencing stay-green and maturity traits in grain sorghum

Trait(mapping method)

QTL LG Marker Position Intervala LOD R2(%) ADDb

Left Right

Stay green SGA A TXS307 16.7 )17.0 14.8 2.8 16.1 0.20(simple) SGD D2 TXS1537 2.9 )4.7 O� end 3.6 16.9 0.17

SGG G UMC27 10.0 )33.2 23.9 2.0 11.1 0.20SGB B pSB115 0.0 )22.9 o� end 1.6 7.7 )0.13SGI.1 I1 pSB134.2 0.0 o� end 14.8 1.8 8.7 )0.14SGI.2 I2 TXS1541 5.7 )12.5 8.3 1.3 9.6 0.12SGJ J TXS713 27.7 )5.7 O� end 1.3 8.6 0.10

Stay green SGA A TXS307 23.3 )9.2 2.0 6.6 28.6 0.27(composite) SGD D2 TXS1537 2.0 )3.9 O� end 5.0 22.5 0.24

SGG G UMC27 10.8 )10.8 10.8 5.8 25.8 0.31SGB B pSB115 2.9 )3.9 O� end 3.0 14.4 )0.15SGI.1 I1 pSB134.2 0.0 )2.0 12.5 3.8 17.6 )0.16SGI.2 I2 TXS1541 2.0 )5.7 10.8 2.9 13.7 0.17SGJ J TXS713 27.7 )17.1 O� end 2.3 11.6 0.14

Maturity DFB B TXS1299 0.0 )3.1 2.9 10.0 40.1 )2.7(simple) DFG G UMC27 11.6 )38.1 21.9 3.1 17.6 )1.9

Maturity DFB B TXS1299 0.0 )2.0 2.0 12.6 47.5 )2.7(composite) DFG G UMC27 14.1 )14.1 12.5 4.4 20.3 )1.9

a The interval was estimated at a LOD fall o� of )1bAdditive e�ect for every allele of Tx430

585

nescence of leaves when exposed to post-¯oweringdrought stress (Rosenow et al. 1983; Wanous et al.1991; Tenkouano et al. 1993; Walulu et al. 1994). Thehigh correlations between stay-green rating and chloro-phyll index values con®rmed the reliability of using vi-sual ratings for evaluation of the stay-green response.Wanous (1991) also reported high correlation betweenvisual rating for stay-green and green leaf area measuredas a percentage of total leaf area.

Mapping of stay-green and maturity traits

Plant responses to water stress are clearly in¯uenced bythe timing and intensity of stress (Ludlow and Muchow1990), which makes the genetic analysis of such traitsmore complicated. Therefore, the pooled average of fourenvironments was used to identify the QTLs in¯uencingstay-green and maturity ratings in response to droughtstress during a speci®c growth phase (post-¯owering).The regression coe�cients in each environment, how-ever, were scrutinized to evaluate the relative importanceof each QTL. Three major and four minor QTLs in¯u-encing the stay-green trait were identi®ed by both simpleand composite interval mapping methods. Similarly, twomajor QTLs in¯uencing the maturity trait were identi-®ed by both methods.

The three major stay-green QTLs contributed to 41%of the phenotypic variability, while the remaining fourminor QTLs contributed an additional 25% to thephenotypic variability. Alleles from B35 contributed tothe stay-green trait in all stay-green QTLs, except fortwo minor QTLs (SGI.1 and SGB). The three majorQTLs (SGA, SGD and SGG) and two minor QTLs(SGB and SGI.2) were consistently identi®ed acrossmost, if not all, environments, while the remaining twominor QTLs had a high QTL ´ ENV interaction

(Table 4). In general, the QTL analysis indicated thatthe genetic control of the stay-green trait in grain sor-ghum is more complex than suggested by the earlier®ndings of Walulu et al. (1994) based on phenotypicanalysis.

Two QTLs (DFB and DFG) in¯uencing the maturityratings were identi®ed by both simple and compositeinterval mapping methods. One QTL on LG B (DFB)alone contributed to 40±47% of the phenotypic vari-ability (Table 3). The regression coe�cients for bothQTLs were consistently signi®cant in all four environ-ments. Both the QTLs in¯uencing maturity corre-sponded to maturity QTLs identi®ed earlier in sorghumand other cereal crops. The location of one QTL, DFB,corresponded to that of the QTL identi®ed in sorghum(Pereira et al. 1994; Lin et al. 1995) and maize, based onmarkers common to LG B (Fig. 1) and LG I reported byChittenden et al. (1994), and the comparative mappingof maturity QTLs by Lin et al. (1995). The other QTL,DFG, corresponded to the QTLs on chromosomes 1 and9 of maize as identi®ed by markers CDO20 and csu077(Lin et al. 1995).

Together all seven QTLs contributed 63±66% ofthe phenotypic variability of the stay-green ratings,which accounted for 88±90% of the genetic variabilitygiven the heritability estimate of 0.72 across fourenvironments. These values clearly demonstrate thecomprehensiveness of our analysis of the genome foridenti®cation of the stay-green QTLs. Oh et al. (1997)have reported three genetic loci controlling leaf se-nescence in Arabidopsis. The two maturity QTLs de-scribed here together contributed 53±57% of thephenotypic variability, which accounted for 59±63% ofthe genetic variability in maturity ratings (H � 0.90).No other QTLs that contributed signi®cantly to thephenotypic variability of stay-green and maturity rat-ings were identi®ed. However, the possibility that ad-

Table 4 Predicted means and regression co-e�cients of QTLs in¯uencing stay-green and maturity traits as identi®ed using compositeinterval mapping

Trait QTL Regression co-e�cientsa

ENV1 ENV2 ENV3 ENV4 POOLED

Stay-green SGA 0.17*** 0.46@ 0.10* 0.29** 0.27@@SGD 0.21@ 0.32** 0.09* 0.29** 0.22@@SGG 0.34@@ 0.41** 0.28@@ 0.20NS 0.30@@SGI.1 )0.05NS )0.14NS )0.17@@ )0.30*** )0.17@SGI.2 0.20@ 0.09NS 0.12** 0.22* 0.16***SGB )0.08NS )0.23** )0.08* )0.16* )0.16@SGJ 0.10* 0.17NS 0.06NS 0.15NS 0.12**

Predictedmean�SE(B35)b

2.45�0.04 2.28�0.10 2.37�0.04 2.71�0.09 2.45�0.04

Maturity DFB )1.7@@ )1.8@@ )3.0@@ )4.2@@ )2.6@@DFG )1.6@ )1.3*** )1.4** )3.2@ )2.0@@

Predictedmean�SE(B35)b

64.6 � 0.28 65.1 � 0.27 64.6 � 0.32 64.1 � 0.53 64.6 � 0.30

a Statistical signi®cance of the regression co-e�cients is indicated asfollows: @@, P ³ LOD5; @, P ³ LOD4; ***, P ³ LOD3; **,P ³ LOD2; *, P ³ LOD1.3; NS, not signi®cant (LOD less than 1.3)

b Predicted means and their standard errors (SE) are for all QTLswith two alleles of B35

586

ditional QTLs exist in the genome that have similare�ects to the QTLs identi®ed cannot be eliminated, asmore putative QTLs were initially identi®ed for boththe traits, which did not additionally contribute to theresidual phenotypic variability after the identi®edQTLs were ®xed.

The identi®cation of QTLs in¯uencing the stay-greentrait not only facilitates marker-assisted breeding for im-proved drought resistance but also provides a goodstarting point for the elucidation of the genetic basis forsenescence phenomenon in general, and for the stay-greentrait in particular. The physiological processes that cul-minate in senescence appear to be diverse and may haveopposite e�ects on the ®nal performance of the plantunder drought stress. Some of the physiological processesthat lead to expression of the stay-green trait, such asprotection against oxidative damage to cell membranes(Noode n 1988b; Pastori and Trippi 1993) and mainte-nance of proper hormonal balance (Van Staden et al.1988)maybe bene®cial for drought resistance, while otherprocesses, such as reduced translocation of photosyn-thates from stem to grain, will not give any selectiveadvantage of stay-green for drought resistance. Theimportance of understanding the genetic basis for ex-pression of complex physiological traits using moleculargenetic tools has recently been emphasized by Prioul et al.(1997). Therefore, the identi®cation of di�erent QTLsin¯uencing the stay-green trait will facilitate the study ofthe physiological processes underlying the development ofthe stay-green trait and their relative importance to plantproductivity under conditions of drought stress. Molec-ular genetic analysis of the delayed leaf senescence mu-tants in Arabidopsis (Nam 1997) has revealed that one ofthe loci is involved in the ethylene signal transductionpathway in leaf senescence (Oh et al. 1997).

Correspondence between QTLs in¯uencing stay-greenand maturity traits

Plant breeders have selected plants for early maturityin order to avoid yield losses under terminal stressenvironments by providing the plants with a morefavorable growing environment during the critical re-productive phase. However, early maturity may have adirect association with early senescence because of itsdirect relationship with faster translocation of stemreserves into reproductive parts of the plant. Whilethis type of physiological mechanism is bene®cial toplant productivity under conditions of terminaldrought stress, premature senescence mechanisms suchas that activated by damage to cell membranes due todrought-induced free radicals (as reviewed byThompson 1988) a�ect its performance under droughtstress. Therefore it is important to establish the geneticrelationship between these two traits and identifyQTLs that in¯uence the stay-green trait but not ma-turity, in order to maximize crop improvement underdrought stress environments.

Phenotypic data revealed a weak but signi®cantnegative correlation (r � )0.22, P £ 0.05) between stay-green and maturity traits (Table 1), indicating that atleast some common QTLs may a�ect both of these traitsand possibly hinting at some common mechanisms. Thedegree of association between the traits has been ob-served to increase with the intensity of drought stress(Rosenow et al. 1983). Overlapping QTLs were identi-®ed for these two traits on LG A. This association be-tween QTLs could be due to linkage of di�erent geneticloci or indicate that the same gene in¯uences both traits.The phenotypic association of these two traits was alsofurther investigated using marker-phenotype correla-tions. Although no stay-green QTL was identi®ed nearthe maturity QTLs in LG B, most of the markers aroundthe QTL DFB showed a signi®cant (P £ 0.1) negativecorrelation with stay-green ratings, suggesting an indi-rect e�ect of maturity QTLs on the phenotyping of thestay-green trait (data not shown). This negative corre-lation and the overlapping of QTLs partly support theobservation that the phenotypic values of RILs had awider distribution than the parental means (Fig. 3) al-though both the traits were mostly contributed by allelesfrom one parent (Table 3).

The QTL analysis of the two traits demonstrates thepossibility of pyramiding the favorable QTLs from boththe traits in order to improve drought resistance inplants. Two major (SGA and SGD) and three minorQTLs (SGI.1, SGI.2 and SGJ) in¯uencing stay-greenwere completely independent of the QTLs in¯uencingmaturity, as no maturity QTL was observed on theseLGs. Similarly, the maturity QTL (DFB) which con-tributed for 40% or more of the phenotypic variabilityin maturity was independent of the stay-green QTLs.The only minor stay-green QTL on LG B maps quite faraway (93 cM) and was contributed by the same parent(TX430) as the maturity QTL (Table 3). Further studiesare needed to elucidate the genetic basis of the associa-tion between the stay-green and maturity QTLs on LGG, and will indicate if these QTLs are tightly linked orshare common genes. In the former case, breaking thelinkage between these QTLs should help in pyramidingthe genes from di�erent parents; in the latter case, one ofthe two QTLs will have to be selected based on theirassociation with productivity, which is the ®nal measureof drought resistance. The molecular genetic analysis ofthe QTLs in¯uencing stay-green and maturity, and theassociation between these two inversely related traits,provides a ®rm basis for further study of the underlyingphysiological mechanisms and demonstrates the possi-bility of improving drought resistance in plants bypyramiding favorable QTLs.

Acknowledgements We would like to thank Dr. G. Hart and Dr.A. Paterson (Texas A and M University), Dr. Ed Coe (Universityof Missouri, Columbia) and Dr. M. Sorrells (Cornell University,Ithaca, N.Y.) for providing the probes for RFLP mapping. Wewould also like to thank Mr. Charles Wood®n (Texas A and MCenter, Lubbock) for his help in conducting ®eld experiments andevaluating the mapping population.

587

References

Bleecker AB, Patterson SE (1997) Last exit: senescence, abscission,and meristem arrest in Arabidopsis. Plant Cell 9:1169±1179

Blum A (1983) Genetic and physiological relationship in plantbreeding for drought resistance. Agr Water Mgmt 7:195±205

Boyer JS (1982) Plant productivity and environment. Science218:443±448

Buchanan-Wollaston V, Ainsworth C (1997) Leaf senescence inBrassica napus: cloning of senescence related genes by sub-tractive hybridisation. Plant Mol Biol 33:821±834

Chittenden LM, Schertz KF, Lin Y-R, Wing RA, Paterson AH(1994) A detailed RFLP map of Sorghum bicolor ´ S pro-pinquum suitable for high-density mapping suggests ancestralduplication of Sorghum chromosomes or chromosomal seg-ments. Theor Appl Genet 87:925±933

Fehr WR (1987) Principles of cultivar development. McGraw-Hill,New York

Feinberg AP, Vogelstein B (1984) A technique for radiolabelingDNA restriction endonuclease fragments to high speci®c ac-tivity. Anal Biochem 137:266

Gri�ths CM, Hosken SE, Oliver D, Chojecki J, Thomas H(1997) Sequencing, expression pattern and RFLP mapping ofa senescence-enhanced cDNA from Zea mays with highhomology to oryzain gamma and aleurain. Plant Mol Biol34:815±821

Jansen RC (1993) Interval mapping of multiple quantitative traitloci. Genetics 135: 205±211

John I, Hackett R, Cooper W, Drake R, Farrell A, Grierson D(1997) Cloning and characterization of tomato leaf senescence-related cDNAs. Plant Mol Biol 33:641±651

Kleber-Janke T, Krupinska K (1997) Isolation of cDNA clones forgenes showing enhanced expression in barley leaves duringdark-induced senescence as well as during senescence under®eld conditions. Planta 203:332±340

Lander ES, Botstein D (1989) Mapping Mendelian factors under-lying quantitative traits using RFLP linkage maps. Genetics121:185±199.

Lander ES, Green P, Abrahamson J, Barlow A, Daly MJ,Lincoln SE, Newburg L (1987) MAPMAKER: an interactivecomputer package for constructing primary genetic linkagemaps of experimental and natural populations. Genomics1:174±181

Lers A, Khalchitski A, Lomaniec E, Burd S, Green PJ (1998) Se-nescence-induced RNases in tomato. Plant Mol Biol 36:439±449

Lin Y-R, Schertz KF, Paterson AH (1995) Comparative analysis ofQTLs a�ecting plant height and maturity across the Poaceae inreference to an interspeci®c sorghum population. Genetics141:391±411

Ludlow MM, Muchow RC (1990) A critical evaluation of traits forimproving crop yields in water-limited environments. AdvAgron 43:107±153

Nam HG (1997) The molecular genetic analysis of leaf senescence.Curr Opin Biotechnol 8:200±207

Noode n LD (1988a) The phenomena of senscence and aging. In:Noode n LD, Leopold AC (eds) Senescence and aging in plants.Academic Press, San Diego, pp 2±50

Noode n LD (1988b) Whole plant senescence. In: Noode n LD,Leopold AC (eds) Senescence and aging in plants. AcademicPress, San Diego, pp 391±439

Oh SA, Park JH, Lee GI, Paek KH, Park SK, Nam HG (1997)Identi®cation of three genetic loci controlling leaf senescence inArabidopsis thaliana. Plant J 12:527±535

Pastori GM, Trippi VS (1993) Antioxidative protection in adrought-resistant maize strain during leaf senescence. PhysiolPlant 87:227±231

Paterson AH, Lander ES, Hewitt JD, Peterson S, Lincoln SE,Tanksley SD (1988) Resolution of quantitative trait loci intoMendelian factors by using a complete linkage map of restric-tion fragment length polymorphisms. Nature 335:721±726

Paterson AH, Damon S, Hewitt JD, Zamir D, Rabinowitch HD,Lincoln SE, Lander ES, Tanksley SD (1991) Mendelian factorsunderlying quantitative traits in tomato: comparison acrossspecies generations and environments. Genetics 127:181±197

Pereira MG, Lee M, Bramel-Cox P, Woodman W, Doebley J,Whitkus R (1994) Construction of a RFLP map in sorghumand comparative mapping in maize. Genome 37:236±243

Prioul JL, Quarrie S, Causse M, de Vienne D (1997) Dissectingcomplex physiological functions through the use of molecularquantitative genetics. J Exp Bot 48:1151±1163

Rosenow DT (1984) Breeding for resistance to root and stalk rots inTexas. In:MughoghoLK (ed) Sorghum root and stalk diseases: acritical review. ICRISAT, Patancheru A.P., India, pp 209±217

Rosenow DT, Clark LE (1981) Drought tolerance in sorghum.Proceedings of the 36th Annual Corn and Sorghum ResearchConference, Chicago, Ill., pp. 18±30

Rosenow DT, Quiseberry JE, Wendt CW, Clark LE (1983)Drought tolerant sorghum and cotton germplasm. Agr WaterMgmt 7:207±222

Saghai-Maroof MA, Soliman KM, Jorgenen RA, Allard RW(1984) Ribosomal DNA spacer-length polymorphisms in bar-ley: mendlian inheritance chromosomal location and popula-tion dynamics. Proc Natl Acad Sci USA 81:8014±8018

SAS Institute (1989) SAS/STAT user's guide Version 6 (4th edn).SAS Institute, Cary, N.C.

Stuber CW, Lincoln SE, Wol� DW, Helentjaris T, Lander ES(1992) Identi®cation of genetic factors contributing to heterosisin a hybrid from two elite maize inbred lines using molecularmarkers. Genetics 132:823±839

Tenkouano AF, Miller R, Frederiksen RA, Rosenow DT (1993)Genetics of nonsenescence and charcoal rot resistance in sor-ghum. Theor Appl Genet 85:644±648

Thomas H, Smart CM (1993) Crops that stay green. Ann Appl Biol123:193±219

Thompson JE (1988) The molecular basis for membrane deterio-ration during senescence. In: Noode n LD, Leopold AC (eds)Senescence and aging in plants. Academic Press, San Diego, pp52±77

Utz HF, Melchinger AE (1996) PLABQTL: a program for com-posite interval mapping of QTL. JQTL 2:1

Van-Staden JE, Cook L, Noode n LD (1988) In: Cytokinins andsenescence. In: Noode n LD, Leopold AC (eds) Senescence andaging in plants. Academic Press, San Diego, pp 281±328

Walulu RS, Rosenow DT, Westner DB, Nguyen HT (1994) Inheri-tance of the stay-green trait in sorghum. Crop Sci 34:970±972

Wanous MK, Miller FR, Rosenow DT (1991) Evaluation of visualrating scales for green leaf retention in sorghum. Crop Sci31:1691±1694

Xu GW, Magill CW, Schertz KF, Hart GH (1994) A RFLP linkagemap of Sorghum bicolor (L.) Moench. Theor Appl Genet89:139±145

588